There is a moment when a service stops being a project and becomes something people depend on. Everything changes at that point, and very little of what changes is taught anywhere. This course is twenty-one hours on exactly that transition.
Observability comes first, built rather than described. Structured logging with correlation identifiers that survive a service boundary. Metrics with cardinality you can afford, instrumented at the four signals that matter. Distributed tracing through a request that touches three services and a queue, with sampling decisions that keep the cost sane.
Alerting is treated as a design problem with a human on the other end. Service level objectives and error budgets, alerting on symptoms rather than causes, burn rate alerts that fire early enough to matter and rarely enough to be trusted, and a ruthless audit of alerts that wake people up and teach them nothing.
The second half is the human system. Running an incident with a defined commander and scribe, communicating during an outage to people who are not engineers, writing a blameless postmortem that produces changes rather than promises, sustainable on-call rotations and handover, chaos and load exercises, and the readiness review a service should pass before it is allowed to carry real traffic. A complete review checklist ships with the course.